Use Slide Headlines to Make the Company Deck’s Argument Visible

A deck can be polished page by page and still not argue anything. Each slide gets its chart, its three bullets, its tidy footer, and nobody reads the sequence straight through. The sequence is where the argument lives, or where it quietly isn’t.

The check is short. Write down the point each page can actually support from the evidence you have. Then read those points in order with every body covered. Duplicates show up. So do conclusions that arrive before their premises, and leaps that depend on something nobody said out loud. Most of the fixes are rewrites, merges, and moves. You can do them without touching the design.

Three moves make up the pass:

1. Write the point before the headline. The point comes from the evidence on and behind the page, not from how good the deck should sound.
2. Read the points with the slides hidden. What does each page establish, and in what order?
3. Keep a real question open. Not every page has to land a conclusion.

This makes the argument you already have visible. It doesn’t make a weak proposition strong, and it doesn’t decide anything for the people in the room.

Write the page’s point before writing its headline

Start with the evidence packet, not the page. Here’s a fictional one to work with, from a small company called Cairn that sells scheduling software to contractors who maintain buildings.

What the evidence actually shows. Four contractor teams ran their work orders through a shared inbox, a spreadsheet, and a whiteboard on the office wall; dispatchers typed the same visit into more than one of them. Cairn built a scheduling flow where a dispatcher moves a visit on the office screen and the technician’s list follows; it covers weekday work, and after-hours callouts still go by phone. Four small teams, all in one region, ran their own scheduling in it for eight weeks and were still running it that way at the end. No comparison to their old process was recorded. Setting each team up took one session of about half a day, run by the two people who built the product, with the contractor’s office manager in the room. The questions and data fixes that followed over the next several weeks were handled as they came up and never written down.

The company, the deck, and every number here are invented for this exercise. The point is what such a packet will and won’t carry.

Take the page the deck labels Pilot. The label is useful navigation. It tells a reader where they are. It does not tell them what the page establishes, and that’s the headline’s job. What the packet supports is narrow: Four teams completed the trial. The tempting version, Teams schedule faster with Cairn, needs a recorded comparison that doesn’t exist. It upgrades completion into business impact, which is the most common way a headline outruns its evidence and the hardest one to notice, because the claim sounds like the obvious meaning of the trial.

So keep both elements and give them separate work. A small label or section marker for wayfinding; a headline that states the point this page establishes. What you’re removing is the label doing the headline’s job, not the label itself.

Two more pages from the same packet show the same discipline. The page called Problem wanted to say Scheduling is broken. What the four offices described is narrower: the same visit is typed into more than one tool. That claim is worth stating plainly, because it’s the thing Cairn’s product actually addresses. The page called Product wanted to say Our scheduling flow, which is a topic again. The bounded version: A dispatcher moves a visit once; the technician’s list follows. That sentence has a subject, an action, and a consequence, and the page next to it can demonstrate all three.

Read the argument with the slides hidden

Now copy the points into a plain sequence and read them the way an audience will meet them: one after another, with no charts to soften the joins.

Here is what Cairn’s deck said before the pass. Headings on the left, what each page argued on the right.

1. Problem — Scheduling is broken.
2. Product — Our scheduling flow.
3. Pilot — The trial succeeded.
4. Benefits — Teams schedule faster.
5. Rollout — Rollout plan, with a timeline: two weeks per team.
6. Next Steps — Next steps.

Read that way, four problems sit in plain view.

Pages 3 and 4 prove the same thing. The trial succeeded and Teams schedule faster are the same conclusion, twice, with the second one pushed further than the packet goes. One page’s worth of evidence is being spent on two pages, and the second spend buys a claim the first couldn’t afford. This is what merging is for. The pilot page keeps the trial’s actual result; the benefits page goes away, its slot freed for something the deck needs.

The transition from page 3 to page 4 depends on unspoken context. It only works if the reader already assumes that teams using scheduling software get faster. Nobody says so. A transition that leans on an assumption is a place where an audience either nods along or stops trusting you, and you don’t get to choose which.

Page 5’s number has no premise anywhere behind it. Where did two weeks per team come from? Nothing measured it. The number was drawn to give the timeline a plausible shape, and it may even have been roughly right — the sessions were half a day, and the questions that trailed after them ran on for weeks. Being wrong isn’t the problem. Being unsourced is. A number that appears because a graphic needed a number will be defended later by someone who assumes it was measured.

Page 6 asks for nothing. Next steps is a list, not a decision. And a list of steps only makes sense if the reader has already agreed with everything before it, including the claim that page 3 already overreached on.

Notice how little of this is about wording. The fixes are structural: narrow page 1, give page 2 a point, merge pages 3 and 4, rebuild page 5 from records instead of from a graphic, and turn page 6 into an ask. Energetic phrasing cannot supply any of it. If the premise isn’t there, the livelier sentence just gets to be wrong faster.

Leave the real question open

Once the duplicate is merged, Cairn has a free page and one thing the packet genuinely doesn’t settle: what a wider rollout costs. Half a day per team is known, for four small teams, with the builders in the room. The weeks after each session weren’t measured. Whether a larger contractor costs the same is unknown.

That’s not a gap to paper over with a confident line. It’s the most useful page in the deck, and it belongs right where the freed slot is, after what’s known and before the ask.

The page can carry a question. What we don’t know: what setup costs when someone other than the builders runs it. That’s a real question, and the deck means it. Compare the other kind: Ready to scale? asked by a page that has already decided the answer is yes, with the rollout schedule on the next slide. That isn’t a question. It’s a conclusion in a costume, and readers who notice it will start checking the other pages for the same trick.

A genuine question does specific work here. It tells the audience which unknown the next stage would resolve, which makes the ask that follows proportionate: not “fund a rollout,” but “pay for two deployments in which somebody writes the setup hours down.” The second ask is smaller and much easier to say yes to, and it’s the one the evidence can carry.

Two cautions. First, an open page doesn’t excuse a vague one. The question still needs a subject, a scope, and a reason it matters, or the audience reads it as throat-clearing. Second, don’t convert the whole deck into questions. A pilot that produced four completed trials should say so. Questions are for the places where you don’t yet know, and their value depends on those places being rare enough to be believable.

Return each revised headline to its evidence

The headline pass is a diagnostic, not a proof. Everything it produces has to go back to the pages, because a claim that reads cleanly in a list can still be unsupported by the chart sitting under it. Here’s Cairn’s sequence after the pass.

1. Problem — The same visit is typed into more than one tool.
2. Product — A dispatcher moves a visit once; the technician’s list follows.
3. Pilot — Four teams scheduled their own weeks in it for eight weeks.
4. Deployment — Set-up took about half a day per team, with the builders doing the work.
5. What we don’t know — What setup costs when someone other than the builders runs it.
6. The ask — Two deployments, with the setup hours written down.

Every page now has a different job. One states the situation, one states what the product does, one reports what happened, one reports what the switch took, one names the unknown, one requests a decision. Read in order, there’s no jump that requires an assumption the deck hasn’t made. Nothing is repeated, and nothing is promised as done that hasn’t happened. Page 6 asks for something the company hasn’t started yet, and it says so in the present tense of a request rather than the past tense of a result.

Now restore a slide. Page 4 is where an appealing headline was most tempting.

Deployment

Set-up took about half a day per team, with the builders doing the work.

• Four sessions, one per team, each about half a day.
• The two people who built the flow ran all four.
• All four teams have fewer than twenty technicians and sit in one region.
• Not on this page: what the weeks after each session cost. Nobody wrote it down.

The headline a writer wants is Set-up takes half a day per team. It’s shorter, it sounds like a fact about the product rather than a fact about four projects, and it would let the next page promise a rollout schedule. The qualifier is what stops it. With the builders doing the work is not hedging for the sake of caution; it names the reason the number may not transfer, and it’s also the reason page 5 needs to exist. Drop it and the deck starts claiming a rate it never measured.

One more check applies to every page that moved. A chart carries a period, a scope, and a footnote. When a page changes position, those pieces travel with it, and a headline written before the move can end up attaching itself to the wrong evidence. Read the restored deck rather than the list.

On the method itself: the presentation approach this pass borrows from builds talks on messages rather than topic labels. The originating project describes that approach on its own site (assertion-evidence.com, recorded as checked in September 2026), and it frames the work around technical presentations. That framing is the bounded support for the label-versus-claim distinction above. It is not evidence about company decks generally, it comes with no experimental study or sample presentation behind it here, and it says nothing about whether a clearer deck wins a budget or a customer.

Which is the honest limit of the whole exercise. A headline-only read tells you where your reasoning is thin, duplicated, or missing a step. It does not tell you whether an audience follows, agrees, or remembers. That takes people in a room, and it takes a version of this pass where you watch where they stop.

The revisions that help most are rarely the impressive ones. Narrowing Scheduling is broken to a visit typed into more than one tool. Deleting a page that repeated its neighbor. Replacing an unsourced number with a sourced one — in the right direction, from the timeline graphic back to the session records. Turning a question mark into a question instead of a conclusion. Nine times out of ten, the smaller headline is the one you can defend, and the deck is better for the trade.

Frequently asked questions

What does a headline-only read reveal, and what does it not reveal?

It reveals where reasoning is thin, duplicated, or missing a step. It does not tell whether an audience follows, agrees, or remembers. That takes people in a room and a version of the pass where you watch where they stop. The method makes the argument you already have visible; it does not make a weak proposition strong or decide anything for the people in the room.

Why keep a page label separate from the headline?

A label is useful navigation: it tells a reader where they are. It does not tell them what the page establishes, which is the headline's job. What should be removed is the label doing the headline's job, not the label itself. In the Cairn example, the Pilot page supports 'Four teams completed the trial'; 'Teams schedule faster' would need a recorded comparison that does not exist.

What structural problems can appear when the points are read with the slides hidden?

Duplicates can appear, as when the trial-succeeded page and the teams-schedule-faster page prove the same thing, with the second overreaching the evidence. A transition can depend on unspoken context. A number can have no premise behind it, as with two weeks per team. And a page can ask for nothing, as with a list of next steps that is not a decision. The fixes are structural: rewrites, merges, moves, and rebuilding from records.

How is a genuine open-question page different from a conclusion in disguise?

A genuine question names what the deck does not yet settle, such as what setup costs when someone other than the builders runs it. It tells the audience which unknown the next stage would resolve, which makes the ask proportionate: not fund a rollout, but pay for two deployments in which somebody writes the setup hours down. 'Ready to scale?' asked by a page that has already decided yes is a conclusion in a costume. An open page still needs a subject, a scope, and a reason it matters, and not every page should become a question.

What does the deployment page's qualifier do, and what limit applies to the method itself?

The headline 'Set-up took about half a day per team, with the builders doing the work' keeps a qualifier that names why the number may not transfer and why the next unknown page exists. Drop it and the deck starts claiming a rate it never measured. When a page moves, its chart's period, scope, and footnote travel with it, so read the restored deck rather than the list. The label-versus-claim distinction borrows from a presentation approach described at assertion-evidence.com, checked in September 2026, framed around technical presentations; that is bounded support, with no experimental study or sample presentation behind it here, and it says nothing about whether a clearer deck wins a budget or a customer.

Further reading

• Positioning, Tagline, and Slide Headline Are Different Writing Jobs — Read first when a headline keeps drifting into a tagline or theme; it separates those writing jobs so the page's local claim stays clear.
• Make a Company Deck Readable Without Flattening the Design — For the return-to-evidence pass, it shows how a qualifier exiled to a footer stops limiting the claim it was meant to change.